Towards Intrinsic Interpretability of Large Language Models:A Survey of Design Principles and Architectures

Yutong Gao, Qinglin Meng, Yuan Zhou +1

While Large Language Models (LLMs) have achieved strong performance across many NLP tasks, their opaque internal mechanisms hinder trustworthiness and safe deployment. Existing sur…